Lending to Small Businesses: The Role of Loan Maturity in Addressing Information Problems *

Transcription

1 Lending to Small Businesses: The Role of Loan Maturity in Addressing Information Problems * Hernán Ortiz Molina Department of Economics University of Maryland ortiz@econ.umd.edu María Fabiana Penas Department of Finance CentER Tilburg University m.penas@uvt.nl Abstract This paper investigates the determinants of the maturity of loans to small businesses with data from the 1993 National Survey of Small Businesses Finances. Our empirical tests are based on the premise that longer maturities exacerbate the consequences of borrower-lender informational asymmetries. Overall, our results suggest that shorter loan maturities can be viewed as a particularly strong type of covenant that mitigates the consequences of the informational opacity typical of small firms. First, we find that firms that pledge collateral obtain longer maturity loans. Second, firm owners with poor personal credit history are granted loans with shorter maturities, which is consistent with creditors forcing more frequent renegotiation of contract terms and closer monitoring of riskier borrowers. Third, we do not find any effect of stronger firmcreditor relationships on loan maturity. First Draft: November 2002 Comments Welcome JEL Classification: G21, G32 Key Words: Loan Maturity, Small Businesses, Relationship lending * We are grateful to Steven Ongena and seminar participants at Vrije Universiteit Amsterdam for their comments and suggestions. Any remaining errors are our own.

2 1. Introduction Lending to Small Businesses: The Role of Loan Maturity in Addressing Information Problems There has been an increased interest by policy makers, regulators and researchers in the role of small banks in providing capital to small firms. Much of this interest stems from the belief that small firms serve as an engine of economic growth and innovation, and the concern that the impressive consolidation of the banking industry 1 will have a negative impact on the contract terms and availability of credit to small firms. Small businesses are often informationally opaque and are financed by private debt markets, as opposed to large businesses that are more informationally transparent, and can therefore access the public debt markets. While recent research on credit availability, borrowing costs, and collateral requirements for small businesses has provided many insights about small business finance [e.g. Petersen and Rajan (1994, 1995), Berger and Udell (1995), Angelini, Di Salvo, and Ferri (1998), Cole (1998), Degryse and Cayseele (2000)], the determinants of the maturity of loans to small private firms have not been explored. We use the 1993 National Survey of Small Businesses Finances to examine the factors affecting the maturity of loans to small businesses. Our empirical tests are based on the premise that longer maturities exacerbate the consequences of borrower-lender informational asymmetries (e.g. borrowers are more able to shift risk, or more likely to be in financial distress at the time of repayment). We investigate three related issues. First, we examine whether other contract terms, like collateral requirements, help mitigate these information problems and influence the lender s willingness to lend with longer maturity. Second, we explore whether creditors use shorter loan maturities to exert control over riskier borrowers, enforcing closer monitoring through frequent renegotiation of the contract terms and preventing borrowers from engaging in asset substitution. Finally, we also explore whether creditors use the valuable information they obtain through ongoing lender-borrower relationships to determine the maturity of the loan contract. 1

3 There is an extensive literature that emphasizes the role of reputation in mitigating the conflicts of interest between lenders and borrowers. It predicts that firms with short track records will begin their reputation acquisition by relying on monitored bank loans, while those with long-standing high credit ratings will borrow from the open market (Diamond (1989, 1991)). The theory of financial intermediaries as delegated monitors applies specifically to small firms that often lack financial records, and are characterized by a high degree of informational opacity. Financial intermediaries gather information through continuous contact with the firm and the entrepreneur in the provision of multiple financial services. This information is then used to adjust contract terms and monitoring strategies. Recent empirical work examines the effect of firm-creditor relationships on the availability of credit, and on some of the contract terms (interest paid and the collateral pledged) of loans granted to small firms. In general, the empirical evidence points to a significant effect of relationships on the access to credit, and on collateral requirements, but the evidence on borrowing costs is mixed. For U.S small firms, Petersen and Rajan (1994) find that relationships increase the availability of credit, but the effect on the price of credit is small. Cole (1998) finds that potential lenders are more likely to extend credit to firms with which they have pre-existing relationships as a source of financial services. Berger and Udell (1995) focus only on bank lines of credit and find that borrowers with longer relationships pay lower interest rates and are less likely to pledge collateral. For Belgian small firms, Degryse and Cayseele (2000) find that the loan rate increases with the duration of the relationship but decreases in the scope of the relationship, that is in the purchase of other information-sensitive products from the lender. They also find that collateral requirements decrease with the duration and increase with the scope of the relationship. Finally, Angelini, Di Salvo, and Ferri (1998) find that long-standing relationships have no significant effect on lending rates for Italian cooperative bank members, but have a positive effect on the access to credit. Our paper contributes to the relationship lending literature by studying the effect of relationships on another contract feature, the maturity of the loan contract. 1 As a result of the mergers and acquisitions occurred during the last two decades in the banking industry, the average bank size tripled, after adjusting for inflation. The number of FDIC insured commercial banks 2

4 The existing theoretical literature on maturity focuses on debt issued by public corporations 2, and it is not completely suited to explain the determinants of the maturity of debt of informationally opaque firms, which are typically bank loans. Although conventional wisdom suggests that financial institutions use loan maturity to address information problems between small firms and banks, and that stronger firmcreditor ties should be associated with longer maturities as informational asymmetries attenuate, we are aware of no theoretical work rationalizing these ideas. In this context, our work is exploratory and it is intended to guide future theoretical research. Our results show that firms that pledge collateral obtain longer maturity loans, which is consistent with collateral mitigating borrower risk-shifting incentives and the higher probability of default associated with longer maturities. Second, we find that firm owners that have been delinquent on personal obligations are granted loans with shorter maturities, providing evidence that banks force more frequent renegotiation when lending to more risky borrowers. These two results provide strong evidence of the role of maturity in addressing information problems. Finally, after controlling for public information, we find that the relationship variables are not significant in explaining the maturity of loans. We use three different proxies for the strength of the relationships: its length, borrowing concentration (proxied by the number of institutions from which the firm borrows), and the scope, that is the use of other informationally intensive financial services from the lender (e.g. checking and savings accounts). Longer firm-creditor relationships, more concentrated borrowing, and the existence of savings accounts with the lender do not seem to be associated with longer maturity loans. We believe that this paper, together with the evidence summarized above, leads to conclude that the effect of relationship variables on contract characteristics is mixed, and therefore not conclusive. This result contrasts with the clear evidence supporting the effect of stronger lender-borrower relationship on credit availability. decreased from 14,434 in 1980 to 8,080 in 2001, with an average of 424 mergers per year. 2 See Ravid (1996) for a survey of this literature. 3

5 The paper proceeds as follows. Section 2 discusses the hypotheses tested in more detail, section 3 describes the data set and the variables used in the analysis, section 4 presents the empirical tests and results, and section 5 concludes. 2. Hypotheses tested Financial intermediaries play a critical role in private markets as information producers who can assess small business quality and address information problems. Intermediaries collect information about the business and design a loan contract on the basis of the financial characteristics of the firm and its owner, as well as the firm s prospects and the associated information problems. In the case of large firms with audited financial statements, financial intermediaries can use formal debt covenants to require the borrower to take or refrain from various actions. These covenants often involve restrictions linked to specific financial ratios, and require the periodic submission of financial information. Berlin and Mester (1992), and Park (2000) show that by giving banks the right to renegotiate or call loans if covenants are violated, covenants enhanced the flexibility and efficiency of financial contracting. However, effective loan covenants generally cannot be imposed on small businesses that do not have audited financial statements, have agreements with employees, customers and suppliers that are kept private, and do not issue publicly traded securities. As a result of this informational opacity, providers of outside finance cannot obtain hard information about these businesses. Berger and Udell (1998) suggest that in these cases, maturity can be viewed as a particularly strong type of covenant. Financial institutions can use short-term loan contracts instead of longer term lending with formal loan covenants, to gain flexibility and control when lending to small firms, by forcing more frequent renegotiation. In this way banks can prevent borrowers from engaging in risk-shifting behavior, and reduce the likelihood of firms becoming financially distressed. The above discussion leads to our first hypothesis: H1: financial intermediaries use shorter maturities to exert control over more informationally opaque, risky borrowers. 4

6 Our second hypothesis refers to the relationship between the collateral pledged and the maturity of the loan. Banks can prevent opportunistic behavior by lending short term and punishing misbehaving borrowers ex post by increasing interest rates on future loans. As the maturity of the loan increases, the incentives for asset substitution increase, therefore banks will rely on other bonding mechanisms such as collateral provisions. This argument leads to our second hypothesis: H2: firms that pledge collateral are able to obtain longer maturity loans. Our third hypothesis about loan maturity relates the role of lender-borrower relationships in mitigating information problems. It is known that lenders obtain valuable information about firm quality through ongoing relationships with borrowers. This soft information arises as a result of repeated interaction by the loan officer and the firm s management, the accumulation of a repayment history, and bank monitoring over previous loans (Stein (2000), Berger et al. (2002)). In addition to this interaction, relationships can be built through the provision of checking accounts, savings accounts, and the use of other financial services that the lender can monitor to generate valuable private information about the firm s condition. This information can then be used to adjust contract terms like interest rate and collateral (Berger and Udell (1995)), but also the maturity of the loan. If lenders address information problems by shortening loan maturities, stronger relationships should attenuate the agency problems that arise from large informational asymmetries, allowing firms to obtain longer term financing. Our third hypothesis is summarized as follows: H3: stronger firm-creditor ties mitigate information problems, allowing firms to obtain loans with longer maturities. To summarize, we hypothesize that lenders use loan maturity to address the information problems typical of small firms. Empirically, we expect to find that proxies for publicly available information/firm s reputation, good credit histories, favorable indicators of financial health and collateral pledges are associated with longer maturities. 5

7 To the extent that strong firm-creditor relationships generate information about borrowers and attenuate information problems, we also expect lenders to lengthen maturities to borrowers with whom they have close ties. 3. The data and variable selection Our data source is the Federal Reserve s 1993 National Survey of Small Business Finance (NSSBF), which provides detailed information about financing practices for a nationally representative sample of small businesses in the U.S. This survey, which was sponsored by the Board of Governors of the Federal Reserve System and the U.S. Small Business Administration, was conducted during and is representative of 4.9 million small businesses. The target population consisted of all for-profit, nonfinancial, nonfarm business enterprises that had fewer than 500 employees and were in operation as of year-end The survey includes demographic information about the owners and the characteristics of the firm, an inventory of firm s use of financial services, firm s income statement and balance sheet data. Furthermore, it has detailed data about the contract features of the most recent loan that the firm has obtained, and information about the recent credit history of the firm and its owners 3. The survey s focus on small firms is ideal for our purposes. Given that many of the firms in our sample do not have formal financial statements, information problems are likely to be severe. This allows us to examine the maturity of loan contracts in a context in which this contract feature could serve as an instrument to address information problems, and test whether lender-borrower relationships can be important in shaping loan contracts to informationally opaque businesses. Our sample consists of 1,673 loans to small businesses granted during , although most loans were obtained in The contract data corresponds to the most recent loan obtained by these firms. Table 1 presents summary statistics of the loan contract features, for the different loan types. Insert Table 1 here 3 For more details about the 1993 NSSBF see Cole and Walken (1995). 6

8 Almost 60% of the loans are lines of credit 4. The contract terms differ markedly according to the loan type. Lines of credit have a median maturity of one year, ranging from one month to 20 years, of which 62% are secured and 59% have a guarantor. The other loan types (capital leases, mortgages, motor vehicle loans, equipment and other miscellaneous loans) differ from credit lines in two main aspects: they have longer maturities, and they are fully collateralized, except for the other miscellaneous loans. The median interest rates charged are not significantly different across loan types, although they tend to be lower for lines of credit, which could be due to the existence of other price terms for lines of credit, specifically fees against the total amount committed, against the unused portion, or both 5. Table 2 describes the variables used in this study, broken down into six main categories: loan contract characteristics, firm characteristics, lender characteristics, governance characteristics, information/relationship characteristics, and other variables. Among the contract characteristics, INTEREST is the interest rate paid on the loan. COLLATERAL is a dummy variable indicating whether the loan was secured. Collateral is generally provided in the form of equipment, real estate, personal assets of the owner, inventory or accounts receivable. GUARANTOR is a dummy variable indicating whether the loan was guaranteed. Guarantees give the lender recourse against the firm s owners for any deficiency in payment. Guarantees differ from collateral in that they do not involve any specific liens. Finally, MATURITY is the length of the loan contract in months. If the loan was granted as a line of credit the maturity is the length of the commitment. In the maturity regression of the next section we control for these other contract characteristics that can affect maturity. Insert Table 2 here Among the firm characteristics we include the following variables. ASSETS is the firm s total assets, DTA is the total debt to assets ratio, and PROFMARG is the profit margin, computed as the percentage of profits over the firm s sales. MSA is a dummy 4 A line of credit allows a business customer to borrow up to a prespecified limit, repay all or a portion of the borrowing and reborrow as necessary until the credit line matures. 7

9 variable that takes the value of one if the firm is located in a metropolitan statistical area, zero otherwise. We also include six industry dummies (CONSTRUCT, MANUFACT, WHOLESALE, RETAIL, SERVICES and OTHIND) which are selfexplanatory. OTHIND, the other industries variable, is the left-out group in all our regressions. If leverage and profitability were useful indicators of financial health, we would expect a negative association between maturity and DTA, and a positive association between maturity and PROFMARG. The industry dummies are included to control for any industry specific factors that could affect the contract terms. We also include an indicator variable for whether the principal owner of the firm belongs to a minority (MINORITY). This characteristic could be relevant to explain loan maturity (and other contract terms) if minorities are discriminated in credit markets as suggested by Blanchflower et al. (1998) and Cavalluzo et al. (1999). We use four dummy variables to control for credit history. OWNDELINQ indicates whether the firm owner has been delinquent on personal obligations in the past, while FIRMDELINQ indicates whether the firm has been delinquent on business obligations. BANKRUPT indicates if the owner of the firm has declared bankruptcy in the past seven years, and JUDGMENT indicates whether there are judgments rendered against the owner. If lenders use shorter maturities to control more risky borrowers, then we would expect negative coefficients on these variables. Among the lender characteristics, we include the following dummy variables. BANK controls for whether the lender is a bank, defined to comprise Credit Unions, Savings Banks, Savings and Loan Associations, and Commercial Banks. NONBANK controls for non-bank financial institutions, which include Finance Companies, Insurance Companies, Brokerage or Mutual Fund Companies, Leasing Companies and Mortgage Banks. All other lender types (OTHLEND) are the left-out group in our regressions 6. These dummies control for variation in typical contract terms across lender types. We also use a dummy variable in the dataset that indicates whether the lender operates in a 5 Boot, Thakor, Udell (1987), Berkovitch and Greenbaum (1991), and Morgan (1993) show that substituting fees for interest rates reduces default risk. 8

10 market where the Herfindhal-Hirschman index of bank deposits concentration is above 1800 (HHI). This variable attempts to control for any possible effects of credit market competition on loan maturity. The governance characteristics include the legal form of the firm: PROPRIET for sole proprietorships (omitted category in regressions), PARTNER for partnerships, SCORP for Subchapter S corporations and CORP for (non-subchapter S) corporations. The governance characteristics are included because different ownership structures may be related to the amount of private information borrowers have, the risks borrowers take, and their ability to shift risk to the lenders. All of these factors are important in the determination of contract characteristics, and could therefore have a direct effect on loan maturity. The variable OWNMG indicates whether the principal owner manages the firm, which could be important if owner-managers have different incentives than employeemanagers regarding risk choices. Finally, the dummy variable FAMILY indicates whether a single family owns at least 50% of firm. The information/relationship characteristics consist of the following variables. FIRMAGE is the age of the firm in years, and LENGTH is the length of the relationship between the lender and the borrower in years 7. NOBORRINST is the number of institutions from which the firm borrows. The variables representing the scope of the relationship are the following three. CHECKING is a dummy variable indicating whether the firm has a checking account with the lender, SAVINGS indicates whether the firm has a savings account with the lender, and OTHFINSERV indicates whether the firm buys other financial services from the bank (such as transaction services, cash management services, credit-related services, brokerage services, or trust services). The variable FIRMAGE is a proxy for the firm s reputation, in the sense that more public information should be available to investors in the case of older firms. Our first hypothesis suggests a positive association between firm age and loan maturity. We use three different measures of the strength of firm-creditor ties. Our first measure is the LENGTH of the relationship. The main difference between FIRMAGE and LENGTH is 6 OTHLEND includes other non-financial institutions, such as Venture Capital Firms or Small Business Investment Companies, other business firm, Family or Other Individuals, Small Business Administration, other government agencies, American Express, and Supplier Firms. 7 In our empirical analysis we use Ln(1+FIRMAGE) and Ln(1+LENGTH). 9

11 that the former reflects public information, while the latter reflects private information that is only revealed to the lender. If lenders obtain valuable information about the borrower during their relationship, they will use this information to better assess the borrower s risk and adjust the terms of the loan contract. We therefore expect a positive association between LENGTH and maturity. Second, we use a measure of borrowing concentration: NOBORRINST. Under the assumption that more concentrated borrowing generates more/better information for the lender, firms with more concentrated borrowing should be able to borrow longer term. Alternatively, these variables might really be a proxy of the firm s quality. Lower quality firms could be credit constrained at their primary lender, and therefore must seek additional financing in other institutions. Therefore NOBORRINST might simply be a proxy for the firm s riskiness or credit quality. Firms typically also have checking accounts, savings accounts and other financial services with their lenders. These services that constitute the scope of the relationship, might also provide the lender with valuable information about the borrower s quality. We therefore expect a positive relationship between maturity and CHECKING, SAVINGS and OTHERFINSERV. The other variables are FIXEDRATE and TERMSTRUCT. FIXEDRATE is a dummy variable equal to one if the interest rate on the loan is fixed, zero if it is floating. We expect this variable to have a negative sign, since the probability of interest rates rising increases with maturity. TERMSTRUCT is the difference between the yield of a 10-year government bond and the yield of a 3-month Treasury bill at the time the loan was made. We include this variable because there is some time variation in our sample. TERMSTRUCT could affect maturity for tax reasons. Brick and Ravid (1985) argue that borrowers prefer long-term debt when the term structure of interest rates is upward sloping because the present value of interest tax shields is highest then. They rely on the expectation hypothesis that, in a context of an upward sloping yield curve, implies higher interest expenses in early years from issuing long-term debt than the expected interest expense from rolling short-term debt. Therefore issuing long-term debt reduces the expected tax liability in a context of an upward sloping yield curve. Conversely, if the term structure is downward sloping, issuing short-term debt increases firm value. This 10

12 argument holds only if firms select their leverage before the debt-maturity structure is chosen. We expect the sign of this variable to be positive as long as the tax effect is significant and TERMSTRUCT has some variation during our sample period. The median of the continuous variables and the sample percentages for dummy variables are reported in Table 3. The firms in the sample have median assets of $ thousands, their median leverage is 0.58 and their median profit margin is 4.23%. The vast majority of these firms are located in MSA (79%). Interestingly, almost 30% of the firms in the sample had been delinquent on either personal or business obligations. Most of the loans are from banks (88%). The median firm age is 12 years, while the median length of the relationship of firms with lenders of their last loan is 5 years. The typical firm borrows from one institution. On average the yield curve has been upward sloping during our sample period. Insert Table 3 here The firms whose most recent loan is a line of credit are substantially larger than those that recently obtained other type of loans (capital leases, mortgages, motor vehicle loans, equipment loans). Most firms that recently obtained credit lines are incorporated (82%), while only 65% of the firms with other loans are incorporated. Firms that obtained credit lines did it mostly from banking institutions (94%), while the share of banks for other loans falls to 79%. Even though firms that obtained other types of loans are smaller in size than firms that obtained lines of credit, they have similar median age and identical median length of their relationship with the financial institution from which they obtained the loan. Only 29% of lines of credit had a fixed interest rate, while this percentage increases to 61% for other loans. 4. The determinants of loan maturity 4.1 Empirical approach Before we turn to the determinants of loan maturity, note that it is important that we control for the other loan contract characteristics that could affect the maturity 11

13 decision (i.e. whether the loan is secured, a guarantor is required, and the interest rate on the loan). However including the contract characteristics as control variables introduces endogeneity, as all loan contract characteristics, including maturity, could be simultaneously determined. We first run an OLS regression of Ln(Maturity) on all exogenous variables (specification 1). This specification should be viewed as the reduced form for maturity. We interpret the coefficients of the included variables as the effects of these variables on maturity, inclusive of any predicted maturity effect of the endogeneous variables that they may imply. We then estimate the following regression that includes all exogenous and endogenous variables: Ln(Maturity of most recent loan) = Other contract characteristics + 2 Firm Characteristics + 3 Lender Characteristics + 4 Governance Characteristics + 5 Information/Relationship Characteristics + 6 Other Variables + 7 Industry Dummies + 8 Region Dummies + 9 Loan Type Dummies Year Dummies +. We first estimate this regression using OLS (specification 2). The coefficients of the test variables should be viewed as the effect of these variables on maturity, excluding their effects through the other contract features. However, as we noted above, these coefficients could suffer from endogeneity. Unfortunately the current theory is not rich enough to describe the determination of the loan contract features, which could be decided simultaneously or could follow a recursive structure. If the decision on maturity is simultaneous to the decision on another contract feature, then not only the coefficients of the contract variables could be biased, but also the coefficients of all other explanatory variables 8. Therefore we estimate the same regression using two-stage least squares (2SLS), where the other contract characteristics are treated as endogenous variables, and the instruments are firm and owner characteristics, and cost of capital variables that do not belong to the original equation (specification 3). 8 This property can be shown easily using the results on partitioned regression in Greene (1997) 12

14 We use the natural logarithm of MATURITY as our dependent variable to mitigate the impact of outliers, but results are similar if we use the original maturity variable. We control for geographical location by including nine region dummies and for year effects (recall that loans were obtained during ). In regressions for all loans, we also include six loan type dummies to control for differences in maturity attributed to different loan types Empirical results Table 4 reports the results of different specifications for the sample of all loans, with heteroskedastic-robust standard errors. The first two columns report OLS estimates. In column 1 the potentially endogenous contract terms have been omitted from the regression, while they are included in column 2. Column 3 of the table corresponds to two-stage least squares (2SLS) estimates. Insert TABLE 4 here The results in the first column of Table 4 correspond to the coefficient estimates of the reduced form. The coefficients of the regression in the second column of Table 4 could also be inconsistent if all contract terms are determined simultaneously. It is important to note that if endogeneity is a severe problem, then all our estimates could be biased, and not only the coefficients of the endogenous right-hand side variables. To address the concerns discussed above, the third column reports 2SLS estimates, where all of the contract terms (INTEREST, COLLATERAL, and GUARANTOR) were treated as endogenous regressors. The vector of instruments includes cost of capital variables, together with firm and owner characteristics that do not belong to the original equation. A more detailed description of the instruments is given in TABLE A0 in the Appendix. The test of overidentifying restrictions reported at the bottom of Table 4, and the first-stage results reported in Table A1 in the Appendix suggest that our instrumentation performs well. Comparing the results in column 2 and column 3 we observe some changes in magnitude of the coefficients, and a few changes in statistical significance. 13

15 The results that are robust to the alternative specifications are the following. The coefficient of COLLATERAL is positive and significant. This result corroborates our second hypothesis that pledging collateral reduces the incentives to shift risk and reduces the lender s exposure in case of payment default, therefore allowing for longerterm borrowing. The coefficient of DTA is negative, which is consistent with lenders shortening maturities to more leveraged borrowers, although it is not significant. Among the credit history variables, FIRMDELINQ does not turn out to be a significant variable, although the sign of the estimated coefficient is negative as expected. Interestingly, the variable indicting whether the owner has been delinquent on personal obligations, OWNDELINQ, is negative and highly significant in all specifications, suggesting that lenders use shorter maturities to exert control over borrowers with poor credit histories, corroborating our first hypothesis. The results on FIRMDELINQ and OWNDELINQ are highly consistent with the recent increase in the use of hard information for smallbusiness lending that relies on the business owner s record in paying off personal debt, as opposed to the less reliable information on the business itself. Lenders such as Wells Fargo are increasingly applying to small businesses the statistics-based methods long used to review consumer applications for credit cards and mortgages. They pinpoint a few pieces of hard information, mainly related to the owner s personal record. Turning to the information/relationship characteristics, firm age does not turn out to be significant, providing evidence against the argument that as firms gain reputation they can obtain longer term financing. Furthermore, in conflict with our third hypothesis, none of the three measures of the strength of firm-creditor relationships appears to be important in explaining maturity either. FIXEDRATE is significantly negative in both OLS specifications. However, we should note that this variable loses significance when we endogeneize the interest rate. One of the instruments used in the 2SLS is FIXEDRATE, which is a significant determinant of the interest rate in the first stage. Therefore we can conclude that the negative effect of FIXEDRATE on maturity is due mainly to its effect on the interest rate. TERMSTRUCT turns out not to be significant. This result is consistent with the empirical literature on the maturity of corporate debt that has found no evidence supporting the tax hypothesis [(Barclay and Smith (1995), and Guedes and Opler (1996)]. 14

16 We should also note that the results regarding the term structure may not be conclusive simply because there is in fact little variation of the yield curve during our sample period. Overall, the results in Table 4 provide support for our first two hypotheses, and indicate that lenders use maturity as a strong type of covenant by forcing negotiation more often when borrowers are riskier or pledge less collateral. However, contrary to our expectations, the results do not provide support for the relationship hypothesis. One possibility is that the lack of significance of the relationship variables is driven by the fact that our sample includes six different loan types, and that for some of these loan types relationships might not be really important due to the nature of the loan. This possibility is investigated below. It has been suggested that the effect of firm-creditor relationships on loan arrangements should be stronger for lines of credit (LCs), because they represent a commitment by the lender to provide future financing under prespecified conditions. Berger and Udell (1995, p. 353) argue that LCs represent a formalization of the firmcreditor relationships and that LCs are more relationship-driven than other loans that are transaction-driven. It is also important to note that, with the exception of lines of credit lines and other miscellaneous loans, the rest of the loans is typically fully collateralized, and therefore the role of relationships may not be important. To investigate whether the lack of significance of the relationship variables could be explained by the reasons discussed above, we examine the role of relationships for LCs separately. Table 5 reports the regression results for lines of credit only (specifications are similar to those in Table 4). The test of overidentifying restrictions and the first-stage results reported in Table A2 in the Appendix again suggest that our instrumentation performs well. Insert TABLE 5 here We again find that pledging collateral increases maturity, and that firm owners that have been delinquent in the past obtain shorter term financing. These results are robust across our different specifications. The coefficient of on the firm age variable is 15

17 positive but either not significant or marginally significant. Furthermore, even though we restrict attention to lines of credit, the relationship variables remain insignificant. Thus, our evidence on loan maturities does not support the presumption that the effects of relationships are stronger for lines of credit than for other loans. To summarize, while our evidence supports the general hypothesis that financial intermediaries use maturity to address information problems when lending to small firms, we do not find evidence that the length or scope of bank-creditor relationships are important in explaining loan maturity Robustness Checks We conduct several robustness checks. First, most of the firm financial characteristics correspond to 1992, while some of the loans where granted in This means that for some firms in our sample the financial data that we use in our regressions as explanatory variables were not available at the time of the loan. Unless the firm s financial condition was fairly stable, this could introduce biases. Of the 1,673 loans in our sample, 1,463 were obtained in We repeated our regression analysis focusing only on loans made after 1993 and our qualitative results remain unchanged. Second, we added dummy variables indicating the primary use of the loan or line of credit (categories were working capital, motor vehicles, other equipment or machinery, leasehold improvements, land and buildings, furniture and fixtures, inventory accumulation, and debt relief). If the proposed use of the funds by the borrower provides the lender with valuable information about risk choices, these variables could affect contract terms. Including these dummies does not affect our results. Third, in the LC analysis, a large proportion of the firms in our sample obtained LCs with maturities of exactly one year. Given the limited variability of our dependent variable in the LCs regressions, we repeated the analysis after excluding loans with maturities equal to one year. Our estimates become less precise. However the results regarding COLLATERAL and OWNDELINQ remain similar to those in Table 5. 16

18 In order to check that the lack of an effect of the relationship variables on maturity is not due to the way we constructed them, we run a logit model to examine the effect of our relationship variables on the probability of a loan being secured. Our results are similar to those of Berger and Udell (1995), who use a similar sample and find that stronger firm-creditor ties lead to a lower probability of pledging collateral. Specifically, we find that firm age, longer relationships, more concentrated borrowings, and the existence of savings accounts with the lender all reduce the probability of pledging collateral. Thus, our relationship variables seem to correctly capture the strength of the relationship, which corroborates our result that there is no effect of relationships on loan maturity. 5. Conclusion Using the 1993 National Survey of Small Businesses Finance, we study the determinants of a loan contract feature, maturity, which has not been analyzed before. Given the lack of a theoretical framework, we estimate different specifications, including a reduced-form, a structural form assuming the other contract characteristics as predetermined, and a structural form correcting for potential endogeneity of the other contract terms. The results that prove to be robust to all specifications are the following. We find that the relation between collateral and maturity is positive and very significant, corroborating the hypothesis that loan maturity increases the pledging of collateral, because both the probability of asset substitution and the probability of default increase with maturity. Our results also show that firm owners that have been delinquent on personal obligations are granted loans with shorter maturities, providing evidence that banks force more frequent renegotiation when lending to more risky borrowers. Both results point to the fact that lenders use the maturity of the loan to address information problems. Finally, after controlling for public information, we do not find that the relationship variables are significant in explaining the maturity. Specifically longer firmcreditor relationships, more concentrated borrowing, and the existence of savings 17

19 accounts with the lender (scope of the relationship) are not found to be associated with longer maturities. While the evidence in the literature points to a strong effect of borrower-lender relationship on the access to credit, our results, together with the evidence summarized in the introduction, leads to conclude that the effect of relationship variables on contract characteristics is mixed, and therefore not conclusive. Further research should address the issue of whether maturity policy changes with the size of the lender. The model of Stein (2002) predicts that small organizations do better than large organizations in the processing of soft information. Berger et al. (2002) test this prediction in the context of bank lending to small firms, and they find that in fact larger banks interact more impersonally with their borrowers, have shorter and less exclusive relationships, and do not alleviate credit constraints as effectively. Their findings raise the issue of whether larger banks might be less willing to lend long-term to small firms. 18

20 References Angelini, P., Di Salvo, R., Ferri, G., Availability and cost of credit for small businesses: Customer relationships and credit cooperatives. Journal of Banking and Finance 22, Barclay, M., Smith, C., The maturity structure of corporate debt. Journal of Finance 50, Berger, A., Udell, F., Relationship lending and lines of credit in small firm finance. Journal of Business 68, Berger, A., Udell, F., The economics of small business finance: the roles of private equity and debt markets in the financial growth cycle. Journal of Banking and Finance 22, Berger, A., Miller, N., Petersen, M., Rajan, R., and Stein, J., Does function follow organizational form? Evidence from the lending practices of large and small banks. Working paper. Berkovitch, E., Greenbaum, S., The loan commitment as an optimal financing contract. Journal of Financial and Quantitative Analysis 26, Berlin, M., Mester, L., Debt covenants and renegotiation. Journal of Financial Intermediation 2, Blanchflower, D. G., Levine, P. B., Zimmerman D. J., Discrimination in the Small Business Credit Market. NBER working paper Boot, A., Thakor, A., and Udell, G., Competition, risk neutrality and loan commitments. Journal of Banking and Finance 11, Brick, I. E., Ravid, S. A., On the relevance of debt maturity structure. Journal of Finance 40, Cavalluzo, K., Cavalluzo, L, Wolken, J., Competition, Small Business Financing, and Discrimination: Evidence From a New Survey. Journal of Business, forthcoming. Cole, R., The importance of relationships to the availability of credit. Journal of Banking and Finance 22, Cole, R. A., Walken, J. D., Sources and uses of financial services by small businesses: Evidence from the 1993 National Survey of Small Business Finances. Federal Reserve Bulletin 81,

21 Degryse, H., Cayseele, P., Relationship lending within a bank-based system: evidence from a European small business data. Journal of Financial Intermediation 9, Diamond, D., Reputation acquisition in debt markets. Journal of Political Economy 97, Diamond, D., Monitoring and Reputation: The choice between bank loans and directly placed debt. Journal of Political Economy 99, Guedes, J., and Opler, T, The determinants of the maturity of corporate debt issues. Journal of Finance 51, Greene, W., Econometric analysis. Prentice Hall, 3 rd. edition. Morgan, D., Financial contracts when costs and returns are private. Journal of Monetary Economics 31, Park, C., Monitoring and structure of debt contracts. Journal of Finance 55, Petersen, M., Rajan, R., The benefits of lending relationships: Evidence from small business data. Journal of Finance 49, Petersen, M., Rajan, R., The effect of credit market competition on lending relationships. The Quarterly Journal of Economics 1109, Ravid, S., Debt maturity, a survey. Financial Markets, Institutions and Instruments, New York University Salomon Center, vol. 5, no. 3. Stein, J., Information production and capital allocation: Decentralized vs. hierarchical firms. Journal of Finance, forthcoming. 20

22 TABLE 1 Maturity and other loan characteristics by loan type Maturity in months % with % with Median Type of Loan # obs Mean Median Min Max Stdev. Collateral Guarantor Interest Rate (%) Lines of Credit 1, % 59.0% 8.0 Capital Leases % 58.0% 8.9 Mortgages % 56.9% 8.5 Motor Vehicle % 39.8% 8.4 Equipment % 50.0% 8.5 Other Misc % 52.6% 8.5 All Loans % 55.9%

23 TABLE 2: Variable Description Variable Name Description Contract characteristics INTEREST Interest paid (%) COLLATERAL =1 if loan is secured, =0 otherwise GUARANTOR =1 if loan is guaranteed, =0 otherwise MATURITY Maturity (months) Firm Characteristics ASSETS Total firm assets (in $) DTA Total Debt/Total assets PROFMARG Profit Margin (% of sales) MSA =1 if the firm is located in a MSA, =0 otherwise CONSTRUCT =1 if firm is in the construction sector, =0 otherwise MANUFACT =1 if firm is in the manufacturing sector, =0 otherwise WHOLESALE =1 if firm is in the wholesale sector, =0 otherwise RETAIL =1 if firm is in the retail sector, =0 otherwise SERVICES =1 if firm is in the services sector, =0 otherwise OTHIND =1 for other industries, =0 otherwise (omitted in regressions) MINORITY =1 if the principal owner belongs to minority, =0 otherwise OWNDELINQ =1 if the owner has been delinquent on personal obligations, =0 otherwise FIRMDELINQ =1 if the firm has been delinquent on business obligations, =0 otherwise BANKRUPT =1 if the owner has declared bankruptcy in the past 7 years, =0 otherwise JUDGMENT =1 if there are judgments rendered against the owner, =0 otherwise Lender Characteristics BANK =1 if the lender is a bank, =0 otherwise 1 NONBANK =1 if the lender is a non bank financial institution =0 otherwise 2 OTHLEND =1 for other lenders, =0 otherwise (omitted in regressions) 3 HHI =1 if the concentration of deposits in the lender s area is high, =0 otherwise Governance characteristics PROPRIET =1 if the firm is a sole proprietorship, =0 otherwise (omitted in regressions) PARTNER =1 if the firm is a partnership, =0 otherwise SCORP =1 if the firm is a subchapter S corporation, =0 otherwise CORP =1 if the firm is a non-subchapter S corporation, =0 otherwise OWNMG =1 if the firm is managed by its owner, =0 otherwise FAMILY =1 if a single family owns at least 50% of firm, =0 otherwise 22

24 TABLE 2: Variable Description (cont.) Variable Name Description Information/relationship characteristics FIRMAGE Firm age in years LENGTH Length of relationship with lender in years NOBORRINST No. of institutions from which the firm borrows CHECKING =1 if firm has checking accounts with lender, =0 otherwise SAVINGS =1 if firm has savings accounts with lender, =0 otherwise OTHERFINSERV =1 if the firm has other financial services with lender, =0 otherwise 3 Other variables TERMSTRUCT FIXEDRATE Term Structure in % points. It is the difference between the yield of a 10-year government bond and the yield of a 3-month T-bill at the time the loan was made. =1 if the interest rate on the loan is fixed, =0 otherwise 1 BANK is defined to comprise Credit Unions, Savings Banks, Savings and Loan Associations, and Commercial Banks. 2 NONBANK is defined to comprise Finance Companies, Insurance Companies, Brokerage or Mutual Fund Companies, Leasing Companies, and Mortgage Banks. 3 OTHLEND includes other non-financial institutions, such as Venture Capital Firms or Small Business Investment Companies, other business firm, Family or Other Individuals, Small Business Administration, other government agencies, American Express, and Supplier Firms. 4 OTHERFINSERV includes transaction services, cash management services, credit related services, brokerage services, or trust services. 23

25 TABLE 3 Summary Statistics for Selected Variables (Medians for continuous variables and sample percentages for dummy variables) Variable All Loans Lines of Credit Other Loans Contract characteristics INTEREST COLLATERAL 72% 62% 87% GUARANTOR 56% 59% 51% MATURITY Firm characteristics ASSETS (000s) DTA PROFMARG MSA 79% 81% 76% CONSTRUCT 12% 11% 13% MANUFACT 17% 22% 10% WHOLESALE 11% 14% 6% RETAIL 20% 18% 23% SERVICES 29% 26% 34% OTHIND 11% 9% 15% MINORITY 9% 9% 10% OWNDELINQ 8.6% 7% 10.8% FIRMDELINQ 20.5% 40% 21.7% BANKRUPT 2% 2% 2% JUDGEMENTS 4% 3% 5% Lender Characteristics BANK 88% 94% 79% NONBANK 10% 5% 17% OTHLEND 2% 1% 4% HHI 51% 50% 53% Governance characteristics PROPRIET 18% 12% 25% PARTNER 7% 6% 10% SCORP 29% 29% 28% CORP 47% 53% 37% OWNMG 77% 74% 82% FAMILY 76% 72% 81% 24

26 TABLE 3(cont.) Summary Statistics for Selected Variables Variable All Loans Lines of Credit Other Loans Information/relationship characteristics FIRMAGE LENGTH NOBORRINST CHECKING 72% 83% 57% SAVINGS 27% 33% 19% OTHFINSERV 45% 57% 28% Other variables TERMSTRUCT FIXEDRATE 41.4% 28.5% 61% Observations